Automatic Detection of Exploitable Vulnerabilities and Patch Generation to Fix the Vulnerabilities

Inspired by the DARPA AI CyberChallenge (https://aicyberchallenge.com), this project focuses on use of Oracle Parfait's vulnerability detector to automatically detect potential vulnerabilities, extract conditions that cause the vulnerability, use those conditions to feed in data to the compiled version of the code using dynamic fuzzing tools or similar, in order to determine potential vulnerabilities that can be exploited in practice.

The second part of the project takes a vulnerability that has been shown to be exploitable, and the conditions that made it exploitable, to automatically generate a patch for the software being analysed to fix the vulnerability through use of LLM techniques.

The project is exploratory in nature. 

Career Level - IC0